Skip to main content
10:00 pm
>> i would hope that this is true. think you very much. >> guest: thank you for having me.
10:01 pm
[applause] i nt9 -- i am james barrat thank you for being here. also to c-span this is such a great bookstore i remember
10:02 pm
when the books were only sold in bookstores and it captures those when i was the kid. but very rich and fascinating. also such a great meeting place. thank you for inviting me. documentary film producer by "national geographic" channel here is some that are available on netflix. with artificial intelligence what i am here to talk about what it is and why i think they think it is developed
10:03 pm
in the wrong way. i really believe that this conversation is the most important of our time. what is artificial intelligence? the theory in development bill it is the subject of study where we tried to penetrate what intelligence is. with neuroscience, intel lot more on top of programming and computer science.
10:04 pm
with him in cognition in the science of a.i. asks what does human to? what are you? what is intelligence? >> with research business sidelight concise that says it is the ability to achieve goals elaborate on dash there's a lot packed into that definition. if it is not doing something that it should be mobile and should come with a body because if you cannot move around and adapt it could be poor quality.
10:05 pm
it comes with the ability is the lever have. they can learn some other animals can learn. i have been interested hover -- in a.i. over several decades by really got it to arrive was working for the learning channel. and i got to interview a pioneer of speech recognition technology that machines that read books to the blind and others has cleaned the terms security that he has reprinted the term that has been around quite a while. now works for google in charge in those research
10:06 pm
iceboat do they think that is the fastest way to create human level intelligence on a machine. that might be something you want to look into. but irobot makes robots also battlefield robots brita there is a debate right now very important whether or not battlefield robots
10:07 pm
should be up front for the kill decision without the human. kind i'd like the drones. ironically i am going to use today and on dash sudan to help us go into a pyramid that has not been explored but has a lot of rock falls. they will be executed even to leave the first what is the fastest way to the burial chamber? what does the lookout look-alike? don't let the title miss lead you but that they were both very optimistic when we share the planet with machines that are smarter than we are.
10:08 pm
and then to help us solve every medical problem with the general mortality. the short a space odyssey before a science-fiction lead -- legend and then went out to win every award about sharing the planet with machines. he said intelligence will win and machines will dominate. that we steer the future not because theophrastus or the stone guest preacher but because you're the most intelligent we share the planet with someone smarter
10:09 pm
than ourselves they will steer the future. that affected me when i started to interview a.i. makers and roboticist shortly after that. to workout this idea when i spoke with artificial intelligence programmers sure enough everybody agreed with the premise most of the decisions affecting our lives will be made by machines. will that transition be friendly? with for a modification will be create machines smarter than ourselves? but i learned if we proceed of the course we are currently following fabled
10:10 pm
develop their own drive. with self protection and quickly we can become their tools. my book is called our final invention is speaks to the need over human intelligence before it is created. with those people that italy is then driven to create smart machines a.i. those that newsday would also it immersed me into the lives to stop the reckless use and development the most intensely enjoyable the also the most harrowing.
10:11 pm
i got more than a bargain for but i found more bad news. so how do we get from smart phones to the machine that threatens our existence? do you think scientists can make a machine as smart as a human? if not been the problem is too hard for something about the human brain that defines the jury. >> the problem is either too hard or something magical or
10:12 pm
mysterious. less than 15% believe the problem is too hard but there is nothing that would be like that. but i do a wider pool of experts and combine them but my conclusion there is nothing magical to create human level intelligence. if you follow that path you may not be aware if you follow that we can crack the problem is systematic time. if intelligence is the
10:13 pm
problem that is when it is about 2045 within our lifetime. with a specialist in non specialist from "the new yorker" magazine he was kind enough to review and said this about how long it would take one century from now and nobody will care. what they care about is what happens next. so what is important but we
10:14 pm
have prepared ourselves. but how exactly how machines get smart enough. there are a couple of chapters about the genius that worked as a code breaker. but what he said i will give you time to read it. i will give you 20 seconds. [laughter]
10:15 pm
i'd like to put it another way we already created machines that are better than we are in other tasks like navigation over the next decade to create machines that are better at research data bang can improve the capabilities. if they jump to vastly super human intelligence but how close are we with the physics experiments and hypophysis for further experimentation so the software system within reach
10:16 pm
fate of the field the of revolutionary algorithms. but the software system that is general intelligence but when that self improving soulful where it exists it could rapidly include its own intelligence then they will have the smarter machine. that takes us back to the question were millions of times more intelligent than we are. what makes us assume they will be friendly? now to switch to real computers like the i am watson is like the technology that i talk about.
10:17 pm
like this machines by the watson team plays out what the computing really is. there massively parallel that means the process a lot currently and not one at a time. it has 30 million pages of common-sense knowledge. using that cognitive architecture with to rating jeopardy champions it is not a trivial game it involves word meanings, knowledge of everything to be the collection of cognitive powers for statistical reasoning but hypophysis
10:18 pm
generation is important regenerate hypotheses all the time we statistically weighed them all the time. how long will this guy talk about artificial intelligence? you could hypothesize 40 minutes. that is the hypophysis. but what said is strange to take licensing exams. to perform medical diagnostics it would be a consulting physician but to license with certain kinds of liability.
10:19 pm
>> you know, the a.i. functions are good when they take our jobs and competes in the job market. by a.i. and automated sports writers and drink -- bank tellers postal workers clerical workers, pharmacists soon to be replaced by medical dietitians right now the average driver is the literacy but when it is risen overall we will be happier then goes the driver's. software developers, a
10:20 pm
recent article of m.i.t. technology the 45% it is a conservative estimate. how close is to the intelligence? it is so close that it is job number what the four companies and governments. what they poured billions of dollars to create virtual brains. that the price of a computer's the most lucrative product of the world. it is the most lucrative product in the history of the world. imagine those working on cancer research, weapons development imagine that product to compete to drive
10:21 pm
the price down. who would not want that technology? this is a short list of the people going for him intelligence. including nsa. artificial general intelligence giving it 1 billion euros to as his project led by gary markham like to go. -- google. think the mind was just acquired by groupe bull for $400 trillion. writing about this for a long time before they became a billionaire said the sale
10:22 pm
is that google's sets up the board for ethics and safety. it is a giant millstone that setting a high bar for future purchases that once these guidelines cannot once they get out but the industry needs guidelines all of us who are thinking about those issues is the great acknowledgement. if google's does not appoint the board and the shareholders google is not afraid of lawsuits but they will have to prove
10:23 pm
themselves in the court of public opinion. what is the one thing these groups have in common at a i will dominate the to do first century. you might ask yourself what could possibly go wrong? makes machines carrying out algorithms. energy transportation, and our banking system and computers are every where how do we jump to dangerous? because of this man. he is the ted maker for understanding artificial super intelligence and i
10:24 pm
have two chapters of the final intervention. he uses the theory of economics. get that humans or machines has preferences of utility functions. to make a predictable when the economist proposed is this that we're not rational all the time you're impulse buyers. but we can probably anticipate that smart machines would be logical of that economic sense. to have a basic drive with self protection and efficiency and increase.
10:25 pm
self improving machines whether that goal is to play chess and to succeed they will need resources or whatever is most expedient. they will also seek to avoid failure like being turned off or unplug. they will protect themselves they will be efficient and squander resources. for creative ways to achieve their goals. since it is one witting route here is the rub. with the risk of a i behavior sue for intelligence is not imply benevolence super
10:26 pm
intelligence is dangerous they will try to overthrow us but it is useful to use all available resources to achieve that goal. we're not talking about syria but sued for intelligence. we have to get used to such a big ideas that in pursuit of the locals there would think to manipulate matter that would solve the problems of nanotechnology. a.i. does not love you or hate you but you are used for something else. it will not share our values by default it will create immense power that but hundred% a moral without extremely careful programming like social
10:27 pm
past. the obvious solution is to give value human life and property. if we cannot tell when life begins how to retell machines how to protect the life? they have fought hard time defining humans from women and children. if we declare we want to be safe if happiness is the coal a powerful machine will stimulate the brain pleasure center. we could argue about what is right and wrong all day and not reach agreement. in addition from culture to culture of the good life
10:28 pm
slave holding is different from the good life today. what will live be one century from now? it does not conflict with your goals with what is best for you. but with the concept of friendliness. these are the concepts win be share the planet with those smart of the demons. it gets worse. before we can figure out how to make friendly machines nsa duty tries to kill humans on the battlefield. if he's killed the machines are friendly the shareholders will be very disappointed. but think about what autonomous really is.
10:29 pm
machines that kill people? sixty-six nations are developing battlefield robots. that will then take an killed humans without a human in the loop by 23040% of combat forces our robot. 2030. not just the autonomous humanlike machine things that carry equipment because 50 percent of the battlefield will be robots. what i hope you get it is really just around the corner this time there really is. and a.g.i. is not all the
10:30 pm
uncontrollable but those who cannot even think about how to control it. it illustrates with the advanced technologies it runs far ahead of stewardship. look at the science of bombs and nuclear reactors and nuclear submarines would it was dreamed about it was no way to split the atom but the more i have learned it was rapid the westernized. in the same way a.i. quickly turned into a weapon over the next 50 years with a gun pointed at its own head. because there's no plan for maintenance of this technology.
10:31 pm
that was a disaster. we did not kill ourselves but we have a gun to our head but that is not the adaptation. but they will fail with that technology that is a.i.. is there a solution? yes. there is a lot on the drawing board. in the 1970's research to have a meeting in california to have basic guidelines you might contaminate the environment. the guidelines are still in place but to keep bad reactions in accidents from happening.
10:32 pm
a.i. has more guidelines the atomic energy agency that cost a lot of money. that without of rule to create such an agency but now i am actually hopeful before we have to suffer an accident to get it. so my sincere thanks for being here tonight. to him glad you have to dig into this conversation i am happy to answer any questions. [applause] >> do you equate self awareness to consciousness or is there a difference?
10:33 pm
>> i think there is a difference for the computer to be self aware it does not need to have consciousness' the way we understand it but to have a model of the environment. and also hast to know added deep level programming. all the nuances of consciousness are beyond us in the short term. we don't know enough about consciousness but everything else that we do hold the other parts of the toolkit i think are more within reach. >> what about the possibility of the programmers are smart but fallible will come up with a technology that just as if
10:34 pm
you turn on the computer it will be infected. had you thought about that? >> not much. a very interesting thought. there will always be saboteurs and mel where. -- malware. we will see more and more a.i. programmers creating bad programs. i worry about that with relationship to our energy grid. but i read the guys who want to stop the a.i. smart enough to create programs to outsmart the programs from those from a lot of money? i think it is possible.
10:35 pm
>> i may have misunderstood but you are looking for two types of safeguards? >> the a.i. community this is about interviews with a.i. developers. we want safeguards. >> but by definition the similarities are rapidly expanding intelligence? >> once you get the super intelligence yes. there is a group the machine intelligence research the is
10:36 pm
trying to have safe a.i. that those programs that creates the lead dna. that tries to learn lessons from the industrial process built in from the inception. right now and advanced cognitive architecture but we will put a condom on its. but you have to start from scratch. there is a book called normal accidents about the development in to learn a lesson for labatt -- from that and gets $20,000 in donations per year with the digit long dash budget of 50 billion.
10:37 pm
>> when is it apparent these machines get the upper hand in to think that the machine is wonderful with that the emotional relationship that somehow that conversation is on their side. and that is all due. >> the answer for more fake bias. all the time we think hot
10:38 pm
that computers relate to us and siri relates to a does have you seen the movie "her"? i saw a robot once at m.i.t. kismet has the aspect of a child when you are with it your eyes get big if it is scared it will shrink back. and all the reactions so complex they could not point to just the complexity problem. >> but it was in the motion
10:39 pm
you get fond of it very quickly. >> and about three-quarters of the way through your book and i find it fascinating if you drill down deeper from a crisis in this species sending crisis, there are a lot of questions raised and i wonder as i was reading it for research to interview what is the nature of consciousness and can it be reduced? >> the chapter that i turn to now with the people that think consciousness is the backbone of intelligence but i said in the book that the
10:40 pm
issue of consciousness is so big and so important with that computational narrow scientist what do you make of this? and now to feel vindicated you're justified i think what is developing is a distinction that i have grasped is product development not research. that it is not the best answer overall. with rapid product research you could discover the consciousness with the offense even more slippery
10:41 pm
so we're doing fast product development. in the 30's and early 40's then be held to ransom for 50 years. >> [inaudible] >> that is a great question. you could not help admire him. but what he thinks but we will create brains faster than stand-alone a.g.i. then
10:42 pm
we are safe. because we will be super intelligence. but that is because we think we are safe. [laughter] than those the point get augmented our soldiers first they are not dumb but they're not pacifists. that is not exactly a recipe for safety either. to using during a.i. research how would they react? >> their reaction we get specific error to humanlike.
10:43 pm
>> i get that way with some people. [laughter] they seem lifelike. [laughter] i think we will experience that. that is the problem that is what we should worry about with music for battlefield and other things. but if we make it that far then we will claus right over it. >> just like the research project, that was when big
10:44 pm
things are created. to go straight from the computer to the printing press. to take up the other stuff you could pick out what and years later they pointed out the head of the science program. the of the individual programmer. with said the discussions looking at the leering of the algorithms.
10:45 pm
doing the coding to inform the intelligence. >> i know something that is very similar to that. this is what shocked me. they already said they would replace us. then they would skip to that. something about our essence to carry on to the future. they accepted that with the personality that is the code that is a fascinating fade with a.i. that is worth preserving the waterfall qualities that we have. that we would be replaced or it is hard to tell who was
10:46 pm
machine or computer. what can we instill with the super intelligence? >> i am not sure how relevant it was but these think religion will be preserved? i know it seems doubtful. >> i don't think so but there is a woman named a and forest a theologian when i saw kismet at m.i.t. righty about that issue. but if we create things with seoul's that we needed a guide for how to treat them
10:47 pm
those guides are the old testament and other text. did they develop spiritual wonder? >>. >> to come across descriptions with those that can take almost all the emotion out and the end result was the same. people like that are very indecisive they cannot make up their minds because the motion gives the drive to do things. karen they replicate the emotion and the drive? >> that is a great question. i think based on what has been written and he is so positive pcs positive outcomes.
10:48 pm
he will see their goal driven one of the definitions of intelligence so they will not lacks the drive but pursuing something in original programming. i am not sure that is connected to the motion. where we might have the basic drive to be satisfied. but his a answer would be the machines are driven. >> do you read with attitude? >> in the book they will
10:49 pm
because they compete for the save resources and ultimately with us not just looking for present threats for acquisition by future threats what is 100 years to a machine? so they will look down what could be 50 years later. >> of first huge move advantage to create the first a.i.. like chess. when it gets to me that developed just unplug. we could on behalf of the nsa. but they will say we can't because china will not and pugo will not because i am will not. because there is such a great product at the end.
10:50 pm
>> [inaudible] there is concern raised by the scientist that grow up. they decided to take a chance. [laughter] >> that is the nature of technology with exploration. they were in a desperate place. those now that china is on the brink. so to address it now before we get to that front tear that it could be here sooner than we think. we should probably wrap up but thank you so much for coming thank you for buying the book in taking time to
10:51 pm
come here tonight. [applause] >> that was fascinating and scary. three do have copies of his book if you would like to take it home and get it signed. thank you. we will have more conversations about this in the next 15 years. [laughter] [applause] [inaudible conversations] [inaudible conversations]

Book Discussion on Our Final Invention
CSPAN March 2, 2014 10:00pm-10:52pm EST

Filmmaker James Barrat discusses his book ``Our Final Invention: Artificial Intelligence and the End of the Human Era.''

TOPIC FREQUENCY Us 7, Google 3, Watson 2, China 2, Nsa 2, Gary Markham 1, Brita 1, Malware 1, A.i. 1, Witting 1, Karen 1, Euros 1, Sudan 1, Syria 1, California 1, Seoul 1, Labatt 1, Kismet 1, Italy 1, Lookout Look-alike 1
Network CSPAN
Duration 00:52:00
Scanned in San Francisco, CA, USA
Source Comcast Cable
Tuner Channel v109
Video Codec mpeg2video
Audio Cocec ac3
Pixel width 704
Pixel height 480
Sponsor Internet Archive
Audio/Visual sound, color

disc Borrow a DVD of this show
info Stream Only
Uploaded by
TV Archive
on 3/3/2014